Download Calculations of Strong Field Multiphoton Processes in Alkali Metal

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Density of states wikipedia , lookup

Renormalization wikipedia , lookup

Hydrogen atom wikipedia , lookup

Time in physics wikipedia , lookup

Quantum electrodynamics wikipedia , lookup

History of quantum field theory wikipedia , lookup

Aharonov–Bohm effect wikipedia , lookup

Yang–Mills theory wikipedia , lookup

Photon polarization wikipedia , lookup

Nuclear structure wikipedia , lookup

Mathematical formulation of the Standard Model wikipedia , lookup

Theoretical and experimental justification for the Schrödinger equation wikipedia , lookup

Introduction to gauge theory wikipedia , lookup

Transcript
Calculations of Strong Field Multiphoton
Processes in Alkali Metal Atoms
K. J. Schafer1, M. B. Gaarde2, K. C. Kulander3,
B. Sheehy4, and L. F. DiMauro4
1
2
Department of Physics and Astronomy, Louisiana State University, Baton Rouge, LA 70803
Department of Physics, Lund Institute of Technology, P. 0. Box 118, S-22100 Lund, Sweden
3
TAMP Group, Lawrence Livermore National Laboratory, Livermore, CA 94551
4
Chemistry Department, Brookhaven National Laboratory, Upton, New York 11973
Abstract. The development of a new class of laser systems, capable of producing
intense radiation in the mid-infrared (MIR) regime (photon energies between 0.3 and
0.4 eV), opens the possibility of observing multiphoton processes in a new class of
systems with lower ionization potentials than those previously studied. Of particular
interest are the alkali metal atoms, which are true one-(valence)-electron systems. We
present theoretical calculations of above threshold ionization (ATI) and high harmonic
generation (HHG) from alkali metal atoms subject to 3-4 /^m laser irradiation. The
ATI calculations, which use a multiple gauge propagation method, show a striking dependence in the production of high-order photoelectrons on the electron-ion potential.
The HHG calculations illustrate the importance of the strong ground-to-first excited
state coupling in multiphoton processes in the alkali metals.
INTRODUCTION
Interest in strong field multiphoton physics continues to flourish, nurtured both
by fundamental discoveries and a growing list of applications. Until recently almost
all of the available high power lasers provided radiation in the visible/near-infrared
region. This has meant that most studies of multiphoton physics have been performed on rather tightly bound systems such as the rare gases [1]. The advent of
a new class of laser systems, capable of producing intense (1-2 TW/cm2) radiation
in the mid-infrared (MIR) regime (wavelengths between 3 and 4 /mi, photon energies between 0.3 and 0.4 eV) [2], opens the possibility of observing multiphoton
processes in a new group of systems with lower ionization potentials than those
previously studied. Of particular interest are the alkali metal atoms which are true
one-(valence)-electron systems, and can thus be modeled very accurately within a
single-active-electron approximation.
There are many similarities between an alkali metal atom interacting with a
strong MIR laser field, and the well-studied case of a rare gas atom subjected to
CP525, Multiphoton Processes: ICOMP VIII, 8th International Conference,
edited by L. F. DiMauro, R. R. Freeman, and K. C. Kulander
© 2000 American Institute of Physics 1-56396-946-7/007$ 17.00
45
a near-IR (825 nm or 1.5 eV) field. Potassium, for instance, has an ionization
potential (Ip) of 4.32 eV, corresponding to ~ 10 mid-IR laser photons, and argon
has an Ip of 15.76 eV, corresponding to 10-12 near-IR photons. Simple scaling
arguments would indicate that the ionization dynamics of potassium interacting
with 3-4 /^m light at intensities of 1012 W/cm2 should be similar to those of argon
interacting with 825 nm light at intensities of 1014 W/cm2. The Keldysh parameter
7 = Jlp/2Up [3], which compares a free electron's cycle averaged kinetic energy (the
ponderomotive energy Up [4]) to 7P, provides a measure of the necessary strength of
this interaction. As 7 becomes less than one the ionization dynamics are dominated
by tunneling and are well within the non-perturbative regime. Potassium atoms
interacting with a 1 TW/cm2, 4 /zm field will have the same Keldysh parameter,
implying similar ionization dynamics, as the well studied case of xenon with 70
TW/cm 2 , 0.8 //m pulses.
Scaling arguments based on any single parameter, however, are only approximately predictive and may overlook important effects. Returning to the comparison between potassium at 3.2 /j,m to argon at 0.8 //m, we note that the separation
between the ground state and the first excited state in terms of the number of photons is a factor of 2 smaller in the alkali metals than it is in the rare gases. There
is accordingly a very strong coupling between the two states in the presence of the
intense field in the alkali metals which is absent in the rare gases. Our calculations
demonstrate that this coupling has a profound effect on the ionization dynamics
and therefore on multiphoton processes such as harmonic generation [5].
In this paper we present an account of the theoretical methods that we have
used to investigate strong field processes in alkali metal atoms using MIR light.
Many of the methods discussed have previously been applied to studies of the rare
gases using visible/near-infrared wavelengths [6]. However, the MIR wavelength
regime is more demanding computationally than the visible, so we have introduced
a multiple gauge propagation method to increase the efficiency of our calculations.
We also present sample results from ATI [7] and HHG calculations [5]. A discussion
of the first strong field MIR experiments in alkali metal atoms can be found in the
paper by Sheehy et al. in this volume [8].
NUMERICAL METHODS
Our calculations of multiphoton processes in the alkali metals are carried out by
direct numerical integration of the time-dependent Schrodinger equation (TDSE)
using the single active electron (SAE) approximation [6]. To calculate HHG spectra
we use the same methods which have been successfully applied to the rare gases.
To make the ATI calculations at MIR wavelengths tractable, we must modify our
approach to more efficiently describe the ionized electron wave function. In this
section we provide a detailed description of our numerical methods.
We begin by expanding the time-dependent state vector in a mixed basis of
discretized radial functions times spherical harmonics,
46
(i)
where ($, (p\l) = Y®(0, </>). Since we restrict ourselves to linear polarization, the m?
quantum number is conserved. For this reason, the m^ label is suppressed in the
following equations. Atomic units (H = me = e = l, c « 137) are used throughout.
We use the semi-empirical pseudo-potentials given in [9] to describe the valence
electron-ion core interaction. These potentials have three parts, a core potential
Vc, a polarization term Vpoi, and a -1/r potential at long range. Vc represents the
shielding of the nuclear charge by the core electrons, as well as the orthogonality
constraints imposed by the exclusion principle. The polarization of the ion charge
cloud is accounted for through dipole and quadrupole potentials proportional to
r~4 and r"6 respectively. To obtain the correct energies and transition matrix
elements, the short range terms Vc and Vpoi are in general dependent on the angular
momentum channel,
(2)
where V? is the sum of the core and polarization terms in a given i channel. For
i > 3 the potential at small r is dominated by the centrifugal term from the kinetic
energy and the short range potentials are irrelevant. Since the pseudopotential
is ^-dependent it is non-local. However, the range over which it is non-local is
restricted to distances where Vc and Vpoi are non-zero, i.e., close to the ion core.
The potential is effectively local for distances greater than about 5-10 au. As
an example, Fig. 1 shows the pseudopotential for potassium in the lowest three
angular momentum channels. We plot the sum of the pseudopotential and the
centrifugal term £(£+ l)/2r 2 . The corresponding hydrogen potentials are shown
for each channel. For distances greater than about 5 au the two potentials agree,
which means that Vc and Vpoi have become negligible.
1=1
1=0
1=2
0.5
I0.0
g -0.5
I
1
3
r(a.u.)
5
3
r (a.u.)
1
3
r (a.u.)
5
FIGURE 1. Solid line: potassium potential for i = 0-2. We plot the sum of the pseudopotential
and the centrifugal term. The corresponding potential for hydrogen is shown as a dashed line.
47
Mixed Gauge Propagation
The choice of electromagnetic gauge has a large influence on the numerical effort
involved in strong field calculations, through both the number of spherical harmonics imax and the time step St required for convergence [10]. In the length gauge the
TDSE in the dipole approximation for an electron in a time- varying field polarized
along the z axis is
i^. = (HQ + e(t)z)il>l
(3)
where HQ is the field- free ion- valence electron Hamiltonian and £(t) is the MIR
laser electric field.
This gauge is preferred near the ion core where the interaction is reasonably
small, even for intense fields. When the electron is ionized, however, it travels far
from the ion core where the interaction grows rapidly. This has two undesirable
consequences. First, the large interaction requires that a small time step be used in
the time integration. Second, the oscillations in the electron's motion induced by
the field, which are on the order of the free electron oscillation amplitude ~ £ /a;2
[11], require the use of a very large number of angular momentum functions.
If we define the vector potential A(t) and an auxiliary function <3>(t) by the
relations
_
dt ~
>
dt ~ 2
'
then for local potentials we can use the gauge transformation
fr(?,t) = e-*A®*+*™1>v(?,t)
(5)
to write the TDSE is the velocity gauge as
,.
(6)
In this gauge the phase factor that describes the free-electron like oscillations
in the ionized electron's wave function is mostly removed, which results in a large
reduction in the number of angular momentum channels needed for convergence.
The interaction is also bounded when the electron travels far from the ion core.
Close to the ion core, however, the interaction A(i]pz can be very large which
again makes the equations hard to integrate. The two interactions are comparable
in strength at a distance from the ion core on the order of the free electron oscillation
amplitude.
We have developed a mixed gauge time propagation technique for solving the
TDSE which uses the length gauge near the ion core and the velocity gauge at large
distances. We switch gauges in a transition zone that is located away from the ion
48
core. This insures that the potential is strictly local in the transition zone (the short
range ^-dependent terms having decayed) and the local gauge transformation Eq. 5
can be used. The mixed gauge approach allows us to use non-local pseudopotentials
near the ion core, which give a superior description of the atomic potentials as
compared to local pseudopotentials, while minimizing the numerical effort.
To implement the mixed gauge time propagation we use two partially overlapping
radial grids. The length gauge grid extends from r = 0 to r/ and the velocity gauge
grid extends from rv to R, which is the maximum grid radius. The grids overlap,
rv < r/, and the midpoint of the transition zone is TQ. At each time step tn we
propagate both wave functions forward in time to tn+i = tn -f St on their separate
grids. Due to the presence of the artificial boundaries, some error accumulates in ^/
near r/ and in i/jv near rv. Before taking the next propagation step we replace the
"bad" part of the wave function on each grid with "good" wave function that we
obtain by using the gauge transformation on the wave function from the other grid.
For instance, after every time step, the portion of -0/ between r0 and rL is replaced
by the gauge- transformed ifrv at the same radial grid points. Typically, r/ = 30 au
and rv = 20 au. The key to this method is that the errors at each propagation step
are localized near the artificial boundaries. This is due to the fact that we use a
finite difference discretization of the TDSE. We turn next to a discussion of these
finite difference equations.
Discretizing the TDSE
In this section we derive a discretized form of the time-dependent Schrodinger
equation in both the length and velocity gauges. The length gauge discretization
has been discussed before in somewhat less detail in [6]. We use a uniform finite
difference grid, however, non-uniform grid versions of these equations have also
been used in the study of Rydberg wave packets [12]. Instead of discretizing the
TDSE directly, we begin by discretizing the action [6,13,14], defined as
)
ti
(7)
where the Lagrangian is
£. = (rl>\i^-T-V-H,\il>).
(8)
HI is the electron-laser interaction in either gauge. We include the centrifugal term
i(i + l)/2r 2 in V, so that T, referred to as Tr in what follows, is the radial kinetic
energy operator. Variation of the action with respect to ^*, -jj^ = 0, for fixed t\
and t%, yields an equation of motion for ty. This is equivalent to requiring that $*
obey the Euler-Lagrange condition
dtd^
49
For a continuous wave function this procedure leads directly to the TDSE for ^.
For a discretized wave function, the procedure leads to a discretized version of
the TDSE that efficiently accounts for the boundary conditions imposed on ?/;,
especially at small r.
The radial dimension rmin = 0 to rmax = Ris divided into Nr equal intervals with
knots at TJ = (j —1/2)Ar {j = l...Nr}. The wave function is known at these points
only. Typically, Ar = 0.25 au. We assume that ^(R) = 0, as well as nM r ) == 0>
and r^ = 0 a t r = 0. With these boundary conditions, we can derive equations
of motions for the coefficients ty\ = ^(r^).
We begin by choosing approximate integration schemes for the expectation values
in the Lagrangian which are second order in the grid spacing Ar. This will lead to
finite difference equations that are accurate to the same order [13]. After integrating
over the angular variables, the remaining integrals in £ are discretized as
: ^r
Ar
Ar
(10)
Vr2
Ar
dr
Ar
(11)
(12)
As can be seen from these expressions, the derivative term is evaluated at the mid
point of each interval, and the potential term is evaluated at the grid points. The
length gauge interaction term is
(13)
a
TV
(14)
where we have made use of
is the 3j coefficient Q =
t+i
(note that c_i = 0).
Substituting these expressions into £ and shifting Indices on the radial summations in accord with the boundary conditions stated above, leads to
1
e=o j=i
2(Ar) 2
50
We make a transformation to the scaled coefficients </>j which are related to ^ by
These satisfy the normalization condition
a
TV/
'-max •' v r
Imposing the Euler-Lagrange condition with respect to ($)*, we arrive at an
equation for the time evolution of (j>{. This equation is
where
(19)
The dimensionless coefficients a and / are
j
"
r? ~ j 2 - l / 4 '
~
2r|
- j 2 - j + i/4'
^o is the field-free Hamiltonian matrix, which is diagonal in the t quantum number
and tridiagonal in the radial index j. Likewise, H\ is diagonal in j and tridiagonal
in £. Far from the r = 0 boundary both OLJ and /3j approach 1, and we recover the
standard second order finite difference equations.
In the velocity gauge, the interaction term is more complicated since it couples
both the radial and the angular coordinates [15]:
dz
The expectation value which enters the Lagrangian is
f
Straightforward discretization of this expression does not yield a Hermitian matrix
so we first integrate the ^_i^- term by parts and rearrange the t indices to get
an equivalent expression:
51
ax
,
r
./
We write the resulting discretized interaction Hamiltonian as the sum of two terms
(21)
(22)
(23)
The first term, weighted by 1/r^, is tridiagonal in the i coordinate. The second
term couples the radial and angular coordinates, and is tridiagonal in both.
Time Propagation
The Hamiltonian in the both the length and velocity gauges consists of two pieces,
HQ, the atomic Hamiltonian, and the interaction piece HI. In both gauges we use
a split-operator expansion of the full short-time propagator which is unitary and
correct to O(St)3.
(24)
The interaction term is evaluated at the midpoint of the time step. The action
of the exponentials on the time-dependent wave function can not be^ calculated
directly due to the non-diagonal nature of the matrices representing HQ and Hj.
We therefore resort to approximations to the full exponentials which are themselves
unitary and correct to the same order in 5t as the split-operator method. The
propagator for HQ is approximated by the Crank-Nicholson form [13]
(25)
The interaction propagator is handled using a 2 x 2 splitting method previously
used by Richardson and others [16] for cartesian grids. We illustrate this first in
the length gauge. In the length gauge, H\ is diagonal in j and we can treat the
wave function at each radial point separately. For each radial index j, the matrix
that must be exponentiated, L J , is tridiagonal in the i index.
= e-
can be split into even and odd pieces D f = Le + L° as shown below:
52
(26)
0 a0
a0 0
(27)
0 a0
a0 0
0 a2
a2 0
0 a3
a3 0
Where the matrix elements are a^ = £(t)ctrj6t/2. The even and odd matrices
consist of 2x2 block diagonal pieces which are easily exponentiated using
exp < —i
cos a —ism a
cos a
0 a
a 0
(28)
A single propagation step in the length gauge is accomplished via
~l [l-iH05t/2]e-iL°6t/2e
Although the matrices Le and L° do not commute, the symmetric placement of the
operators makes this approximate propagator unitary and correct to O(St)3.
Propagation in the velocity gauge is handled in a very similar manner. Referring
to Eq. 21, we use the symmetric splitting
~l [l-iH06t/2]e~iflv
(29)
The terms involving HI (Eq. 22) are of the same form as the interaction term in
the length gauge and are treated in the same way.
e-iHiSt/2^ =
eMSfl
(3Q)
where the non-zero elements of the antisymmetric matrix of M-7 are
at = A(t)r~l(t+l)ce6t/2. Again the full matrix is split into even and odd 2x2
block-diagonal matrices and the exponentiation is accomplished using the relation
exp
0 a
-a 0
cos a sin a
— sin a cos a
(31)
Exponentiating the terms involving #2 (Eq. 23) is more complicated since they
couple $ to the four adjacent coefficients $±J. To do it, we use the 2x2 splitting
53
method twice in succession, first in the i dimension and then in the j dimension.
The exponent —iH^t/l can be written as a super-matrix, N, which is tridiagonal
0 a0 0
do 0 a\
0 ai 0
N = -iH2St/2 =
(32)
Each of the elements at are themselves matrices that are tridiagonal in j:
at = ,2Ar
2
0
ai
0
—Oi\
0
QL<2
0
-a2
0
(33)
with QJ given by Eq. 20. We now split the super-matrix N into even and odd pieces
which consist of 2 x 2 blocks just as in Eq. 28.
JV
C-
e
0
e^ eN
—— C
(OA\
O
\*J^tl
Before applying each of these exponentials, we diagonalize using a similarity transformation:
_
V2
1 -1
exp
-
0 at
at 0
0
l-lv/2
The application of the operators e±d£ is then made by further splitting the matrices
at into even and odd pieces in the j index as in Eq. 28, and then using Eq. 31.
The various even and odd matrices are placed symmetrically in Eq. 29 so that it
remains unitary.
Both the length and velocity gauge propagators are unitary and unconditionally
stable. All of the various operations scale as Nrx£max, which means that the overall
numerical effort involved in using our multiple gauge approach is linear in the total
number of grid points. We can determine about 1.2 x 106 space-time points per
CPU second on a Cray Y/MP.
RESULTS
In this section we discuss some illustrative results from our calculations of photoelectron and photoemission spectra from alkali metal atoms in MIR fields. Further
details of the theoretical results, and their connection to experiments on alkali metal
atoms, can be found in references [5,7,8].
54
5
10
Energy (eV)
15
FIGURE 2. ATI spectra for potassium (upper curve) and sodium (lower curve) at 1012 W/cm2
and 3.2 /j,m. The ponderomotive energy is 1 eV.
In Fig. 2 we show total (angle integrated) ATI spectra for potassium and sodium,
at a wavelength of 3.2 /mi, corresponding to a photon energy of 0.39 eV. lonization of K requires a minimum of 11 photons at this wavelength. The laser pulse
has a trapezoidal shape, consisting of one cycle ramps and five cycles of constant
intensity. The spectrum is calculated using an energy window function [17]. The
peak intensity is 1012 W/cm 2 , which corresponds to a ponderomotive energy of
about 1 eV. The difference in the ionization yields is due to the higher ionization
potential of sodium as compared with potassium. These results are representative
of calculations over a range of laser intensities. The calculation used tmax = 16 and
6t = 0.1 au. Calculations using the length gauge for the entire radial grid required
(-max = 200 and 8t = .02 au for a similar level of convergence.
Fig. 2 illustrates a striking phenomena in the alkali metals. The production
of high order photoelectrons is very sensitive to the atomic species. K exhibits
a plateau extending up to the about 10C/P, whereas the Na distribution shows
essentially no change of slope beyond 3UP. Fig. 3 shows that this difference persists
over a range of intensities. Indeed, Na never shows any significant high-order ATI
production while K shows an efficiency that is higher than that observed in the
rare gases [18]. Since the high energy portion of the ATI spectrum results from
the rescattering of the electron from the ion core after it has been initially freed
and accelerated in the strong field [18-21], it is reasonable to try to connect this
difference in rescattering efficiency to the electron-ion potential. Our calculations
show that the very efficient rescattering in K is mostly due to the attractive short
range part of the i — 2 potential. This attractive piece is absent in Na. More details
of these results, as well as experimental confirmation of the predictions implicit in
Figs. 2-3, can be found in the paper by Sheehy, et al in this volume and in [7].
One of the most compelling reasons to study strong field processes in the MIR
is the possibility of generating high order harmonics in the UV/visible portion of
the spectrum [5,21]. Potentially these harmonics could be completely characterized (both amplitude and phase) using conventional techniques. This would allow
for a deeper understanding of HHG in general, and may provide insight into the
55
10-3
1
CS
10'5
0
5
10
15
0
5
Energy/Up
10
15
Energy/Up
FIGURE 3. Normalized partial rates (ionization rate per ATI peak) for K and Na for three
intensities between 0.75 and 1.5 TW/cm 2 . The individual points have been scaled so that the
sum of the points for each data set is equal to one. The lowest intensity data set is the top one
for each atom. Note that the energy has been scaled by Up
question of whether the XUV harmonics generated from the rare gases can be used
to produce attosecond pulses. A discussion of the first experimental observation of
HHG in alkali metal atoms (K and Rb) can be found in references [5] and [8].
Fig. 4 shows a single-atom harmonic spectrum calculated for a potassium atom
in a 3.2 /mi, 80 cycle laser pulse with a peak intensity of 1 TW/cm2. The spectrum
exhibits many similarities with the better-known near-IR excitation of rare gas
atoms, especially the general form of the spectrum and the cut off. However, as
mentioned above, the alkali metals present several interesting and potentially useful
contrasts to the rare gases. Chief among these is the strong coupling between the
ns ground state and the first excited up state.
As an example of the role played by this strong coupling, we compare the full
result, shown in white squares, to two approximate calculations. The approximate
calculations start by defining a subspace projection operator Pn:
The projected wave function at a time t is |$n(0) = Ail^WX where \^(t)) is the
full time-dependent wave function. The approximate dipole moment is'calculated
by requiring that the transition either begin or end in the projected subspace:
(1>(t)\Pnz + zPn\<l>(t)) -
(36)
As n —> oo and Pn —> 1, the approximate dipole goes over to the full dipole, as it
should. The last term makes up for double counting of transitions between states
in the projected subspace. An equivalent way to write the approximate dipole that
makes this clear is
(dn)(t)
=
(37)
This applies equally well when we calculate the acceleration form of the dipole.
56
11
15
17
Harmonic (of 3.2 ^m)
21
FIGURE 4. Single atom harmonic spectrum for K. The full calculation is compared to approximate dipoles calculated using the continuum wave packet plus the 4s and the 4s + 4p states.
For potassium, which has a 4s ground state, we compare approximate dipoles
calculated using
(38)
and
(39)
The black squares in Fig. 4 are the result of including the dipole interaction between
the time-dependent wave function and the ground (4s) state. This approximation,
which works very well in the rare gases, fails here, showing differences in both the
conversion efficiency and the harmonic cutoff energy. However, when we include
the first excited state in the projected subspace, as shown with black circles (inside
the white squares), we get perfect agreement with the full result.
The important role played by the coupling of the first excited np state in HHG
in the alkali metals provides an interesting and potentially useful contrast with
HHG from rare gases. For example, it might be possible to achieve control over
the HHG process by using a second laser field to alter either the ns —> np or
the np —» continuum coupling. This is just one example of the potential new
applications that may result from extending strong field multiphoton studies to the
mid-infrared wavelength regime.
ACKNOWLEDGMENTS
K. J. S. and M. B. G. acknowledge support from the National Science Foundation through grant No. PHY-9733890, the Louisiana State Board of Regents
through grant No. LEQSF96-99-RD-A-14, and the Swedish National Science Research Council. Computer time was provided by the National Supercomputer Centre in Sweden. This research was carried out in part under the auspices of the
57
U. S. Department of Energy at Lawrence Livermore National Laboratory under
contract No. W-7405-ENG-48, and in part at Brookhaven National Laboratory under contract No. DE-AC02-98CH10886, with the support of its Division of Chemical
Sciences, Office of Basic Energy Sciences and BNL/LDRD No. 99-56.
REFERENCES
1.
2.
3.
4.
B. Sheehy and L. F. DiMauro, Ann. Rev. Phys. Chem. 47, 463 (1996).
B. Sheehy, J. D. D. Martin and L. F. DiMauro, J. Opt. Soc. Am. B submitted (1999).
L. V. Keldysh, Sov. Phys. JETP 20, 1307 (1965).
Up measured in eV is given by 0.093/A2 where / is the intensity in TW/cm2 and A
is the wavelength in microns. The long wavelength, high intensity limit is defined by
Up being substantially larger than the photon energy. For 1 TW/cm2, 3.2 ^m light
Up is 1 eV, which corresponds to .about three MIR photons.
5. B. Sheehy, J. D. D. Martin, L. F. DiMauro, P. Agostini, K. J. Schafer, M. B. Gaarde,
and K. C. Kulander, Phys. Rev. Lett. 83, 5270 (1999).
6. K. C. Kulander, K. J. Schafer, and J. L. Krause, in Atoms in Intense Radiation
Fields, Ed. M. Gavrila (Academic Press, New York, 1992).
7. M. B. Gaarde, K. J. Schafer, K. C. Kulander, B. Sheehy, D. W. Kirn, and L. F. DiMauro, Phys. Rev. Lett, submitted (1999).
8. B. Sheehy, J. D. D. Martin, T. Clatterbuck, D. W. Kim, L. F. DiMauro, P. Agostini,
K. J. Schafer, M. B. Gaarde, and K. C. Kulander, this volume.
9. W. J. Stevens et a/, Can. J. Chem. 70, 612 (1992).
10. E. Cormier and P. Lambropoulos, J. Phys. B 29, 1667 (1996); H. G. Muller, Laser
Physics 9, 138 (1999).
11. The free electron oscillation amplitude measured in atomic units is given by 2.6\/7A2,
where / is the intensity in TW/cm2 and A is the wavelength in microns. For 1
TW/cm2, 3.2 /mi light the amplitude is « 26 au.
12. J. L. Krause and K. J. Schafer, J. Phys. Chem. A 103 10118 (1999).
13. S. E. Koonin, Computational Physics, Menlo Park, CA: Benjamin/Cummings, Inc.,
1986.
14. J. Negele, Rev. Mod. Phys. 54 913' (1982).
15. J. L. Krause, J. D. Morgan, and R. S. Berry, Phys. Rev. A 35 3189 (1987).
16. J. L. Richardson, Comp. Phys. Comm. 63 84 (1991); H. DeRaedt, Comp. Phys. Report 1 1 (1987).
17. K. J. Schafer Comp. Phys. Comm. 63 427 (1991).
18. G. G. Paulus et a/, Phys. Rev. Lett. 72, 2851 (1994).
19. K. J. Schafer et a/, Phys. Rev. Lett. 70, 1599 (1993).
20. P. B. Corkum, Phys. Rev. Lett. 71, 1994 (1993).
21. B. Walker et a/, Phys. Rev. Lett. 77, 5031 (1996).
58